Sample Selection Bias in Adult Mortality Estimates from Mobile Phone Surveys. Evidence from 25 Low- and Middle-Income Countries

Sahar M A Ahmed, Centre d'Estudis Demografics
Julio E. Romero-Prieto, London School of Hygiene and Tropical Medicine (LSHTM)
David P Sanchez-Paez, Université catholique de Louvain
Bruno Masquelier, Louvain University (UCL)
Thomas Pullum, Demographic and Health Surveys Program
Georges Reniers, London School of Hygiene and Tropical Medicine (LSHTM)

In low- and middle-income countries, mobile phone surveys for adult mortality estimates from sibling survival histories face potential sample selection bias due to non-universal mobile phone ownership (MPO). This study assessed MPO-related bias using 25 Demographic and Health Surveys. It involved logistic regression to examine MPO's link with socio-demographic characteristics, estimating adult mortality probability (45q15) among women of reproductive age, and correcting bias in mortality estimates from mobile phone owners using post-stratification weighting. Results showed MPO's predictable correlation with socio-demographic traits. In 20 out of 25 countries, mortality estimates based on mobile phone owners closely matched the general population, with significant biases found in Papua New Guinea, Burundi, Rwanda, Haiti, and Zimbabwe, which post-stratification weighting effectively reduced. The study concludes that MPO-related selection bias in mortality estimates is generally modest and can be mitigated through post-stratification weighting.